The Caenorhabditis elegans gene lin-26 is required to specify the fates of hypodermal cells and encodes a presumptive zinc-finger transcription factor

How do general categories of cell types, such as epidermal cells, neurons and muscle cells acquire their identities? In vertebrates, a family of cell-type-specific basic helix-loop-helix (b-HLH) transcription factors including MyoD, Myf-5, herculin and myogenin, controls the muscle cell fate in vitro (Weintraub et al., 1991) and in vivo (Hasty et al., 1993; Nabeshima et al., 1993; Rudnicki et al., 1993). In Drosophila, a family of four b-HLH transcription factors, the proneural genes of the achaete-scute complex, specifies the neural cell fate (Ghysen and Dambly-Chaudière, 1988; Rodriguez et al., 1990).

We have chosen to study the problem of how cell fates are specified in the nematode Caenorhabditis elegans. We have focused on the gene lin-26, which was defined by the mutation n156; this mutation causes neural cells to be generated instead of certain hypodermal cells (Ferguson et al., 1987). (In C. elegans, epidermal-like cells are termed hypodermal cells.) Specifically, in wild-type C. elegans first stage (L1) larvae, each of the 12 postembryonic P ectoblasts divides to generate an anterior daughter, the Pn.a cell, and a posterior daughter, the Pn.p cell (Sulston and Horvitz, 1977). The Pn.a cells are neuroblasts that undergo three rounds of cell divisions and generate up to five motor neurons. The Pn.p cells are hypo-dermoblasts that generate the vulva, which is used for egg laying, and other cells of the ventral hypodermis. In lin-26(n156) animals, the Pn.p cells generate neuronal descendants (Ferguson et al., 1987), causing these animals to lack a vulva. Indeed, the mutation lin-26(n156) was originally identified in a screen for vulvaless mutants (Ferguson and Horvitz, 1985).

In the present study, we characterize the role of lin-26 during C. elegans hypodermal development. We have isolated three new lin-26 alleles that we show to cause embryonic lethality and defects in all categories of hypodermal cells. The lin-26 transcript encodes a presumptive transcription factor with two zinc-finger motifs. Based on these observations, we postulate that lin-26 acts as a transcriptional activator to specify the fates of hypodermal cells.

Worms were maintained as described by Brenner (1974). Animals were raised at 20°C unless otherwise stated. This paper conforms to the standard C. elegans genetic nomenclature described by Horvitz et al. (1979). C. elegans strain N2 (Brenner, 1974) was the wild-type strain and the parent of all C. elegans strains that we used. The following mutant chromosomes and mutant strains were used: mnC1[dpy-10(e128) unc-52(e444)] (Herman, 1978); dpy-2(e489) and unc-4(e120) (Brenner, 1974); him-5(e1490) (Hodgkin and Brenner, 1977); lin-26(n156) (Ferguson and Horvitz, 1985); MT5823: lin-26(n156) unc-4(e120)/mnC1[dpy-10(e128) unc-52(e444)]; him-5(e1490); SP754: mnDf88 unc-4(e120)/mnC1[dpy-10(e128) unc-52(e444)], SP781: mnDf97 unc-4(e120)/mnC1[dpy-10(e128) unc-52(e444)], SP803: mnDf105 unc-4(e120)/ mnC1[dpy-10(e128) unc-52(e444)], SP804: mnDf106 unc-4(e120)/ mnC1[dpy-10(e128) unc-52(e444)] (Herman, 1978).

lin-26(n156)/Df animals are inviable (Ferguson and Horvitz, 1985). We found that the mutation lin-26(n156) causes larval lethality in trans to deficiencies that delete lin-26. Hemizygous lin-26(n156)/Df larvae were recognized as homozygous Unc-4 larvae in crosses between MT5823 males and SP754 hermaphrodites (although lin-26(n156)/Df larvae die, most reach a stage at which the Unc-4 phenotype can be observed, which is the late L1 stage; see Miller et al. 1992). Mothers were allowed to lay eggs for half a day and progeny were examined after 40 hours.

Because hemizygosity for lin-26(n156) causes larval lethality, we designed a screen that would allow us to recover lethal alleles of lin-26. The progeny of 4,400 F₁ progeny of dpy-2(e489) unc-4(e120)/mnC1 animals mutagenized with ethyl methane sulfonate (Brenner, 1974) were examined for the presence of Dpy Unc animals. The F₂s from individuals that did not segregate Dpy Unc offspring were tested for complementation with the deficiency mnDf88, which deletes lin-26 (see below), and for complementation with lin-26(n156). Two mutations, named lin-26(mc1) and lin-26(mc2), failed to complement lin-26(n156). The allele lin-26(mc4) was isolated by a similar strategy: the F₁ progeny of 3700 dpy-2(e489) unc-4(e120)/mnC1 animals mutagenized with ethyl methanesulfonate raised at 22.5°C was directly examined for complementation with lin-26(mc1). They were outcrossed twice with N2 and balanced by mnC1 to maintain stocks. To examine the terminal phenotypes of lin-26(mc1), lin-26(mc2) or lin-26(mc4) embryos, heterozygous dpy-2(e489) lin-26(mcx) unc-4(e120)/mnC1 mothers carrying the corre-sponding lin-26(mcx) mutation (where mcx can be mc1, mc2 or mc4) were allowed to lay eggs for 3 hours; the progeny were examined 14 hours later. Hemizygous lin-26(mcx) embryos were obtained by crossing heterozygous mnDf88 unc-4(e120)/+ males with dpy-2(e489) lin-26(mcx) unc-4(e120)/mnC1 hermaphrodites. Heteroallelic lin-26(mcx)/lin-26(mcy) embryos were obtained by crossing dpy-2(e489) lin-26(mcx) unc-4(e120)/mnC1 hermaphrodites with dpy-2(e489) lin-26(mcy) unc-4(e120)/mnC1; him-5(e1490) males, except lin-26(mc1)/lin-26(mc4) embryos, which were obtained by crossing dpy-2(e489) lin-26(mc1) unc-4(e120)/ mnC1 hermaphrodites with dpy-2(e489) lin-26(mc4) unc-4(e120)/+ males. For each of the crosses described above, we checked that the proportion of males was as expected in a successful cross.

The map position of lin-26 on LGII (Ferguson and Horvitz, 1985) was refined by mapping lin-26(n156) relative to deficiencies. MT5823 males were crossed with SP754, SP781, SP803 and SP804 hermaphrodites. mnDf106 complemented lin-26(n156) (36 Unc males and 129 wild-type males), mnDf97 partially complemented lin-26(n156) (two Unc males and 66 wild-type males), mnDf88 and mnDf105 failed to complement lin-26(n156) (no Unc males and over 100 wild-type males). To identify the left breakpoints of these four deficiencies, DNA from strains SP754, SP781, SP803, SP804 and N2 were analyzed by Southern blotting. For probes, we used cosmids from the region of the previously cloned LGII genes unc-104 and zyg-11 (Coulson et al., 1988). The cosmid C05B9 identified the left break-point of the deficiencies mnDf88 and mnDf105, the cosmid C24G10 identified the left breakpoint of the deficiency mnDf97 and the cosmid C01D8 identified the left breakpoint of the deficiency mnDf106 (data not shown).

Cosmid DNA was microinjected into the gonad of lin-26(n156) animals at 10 ng/μl together with the plasmid pRF4, which carries the dominant mutation rol-6(su1006), at 80 ng/μl (Mello et al., 1992). F₁ rollers and stable F₂ roller lines were isolated, and animals were examined for their abilities to form a vulva and lay eggs. (lin-26(n156) animals never lay eggs.) lin-26 was located by testing cosmids surrounding C24G10 for their abilities to rescue the egg-laying defect of lin-26(n156) animals. The cosmids K06E5, C18C9, F31B8 and F18A1 rescued lin-26(n156), whereas the cosmids C36F4, C43A1, C18D9, C24G10, C01B9 did not rescue lin-26(n156). A 17 kb MluI-SpeI fragment from the cosmid K06E5 was subcloned into the plasmid pIBI20 (IBI) (plasmid pMLW003). A 9 kb BamHI and an 11 kb HindIII fragment from pMLW003 were subcloned into the plasmid pBSKII (Stratagene) (plasmids pMLW006 and pMLW009, respectively). Plasmids pMLW003 and pMLW006 rescued the phenotypes caused by the mutations lin-26(n156), lin-26(mc1) and lin-26(mc2). Two deletions were made from pMLW006 by digestion with SacI or SphI and religation. Frameshift mutations within pMLW006 were made by partial digestion with NdeI or EcoRI followed by end filling with Klenow enzyme; as a result the EcoRI site (G/AATTC) was transformed into an AseI site (AT/TAAT). DNA manipulations were performed as decribed by Sambrook et al. (1988).

The DNA sequence of both strands of the region bracketed by an SphI site on the left and a HindIII site on the right (see Fig. 4) were determined using the dideoxy chain termination method and Sequenase enzyme (US Biochemical). Single-stranded templates were prepared using the phagemid pBSKII from deletion derivatives generated by limited ExoIII digestion (Sambrook et al., 1988). Northern blots were performed using N2 or lin-26(n156) mixed-stage polyadenylated RNA, or N2 embryonic polyadenylated RNA. 500,000 plaques from a mixed-stage population cDNA library (Kim and Horvitz, 1990) were screened using the plasmid pMLW006 as a probe. Fifteen cDNAs were analyzed. Fourteen of them, ranging in size between 1.0 kb and 2.2 kb, corresponded to nearly full length or partial cDNAs for the 2.2 kb RNA; the sequences of the longest two (1.8 kb and 2.2 kb) were determined and proved to encode identical proteins (data not shown). The fifteenth corresponded to a partial 0.7 kb cDNA for the 1.5 kb RNA. Further cDNA clones for the 1.4/1.45/1.5 kb RNAs were generated using reverse transcriptase and the polymerase chain reaction (RT-PCR; see Sambrook et al., 1988) using oligonucleotides corresponding to the most 5′ and 3′ sequences of the 0.7 kb cDNA or from sequences that we tentatively predicted to be coding after inspection of putative splice donor and acceptor sites in the genomic sequence (P. Dufourcq, M. Labouesse and H. R. Horvitz, unpublished observations; a detailed account of our strategy will be presented elsewhere). These cDNAs revealed that the 1.4 kb, 1.45 kb and 1.5 kb RNAs differ by alternative exons in their 3′ ends and so generate proteins that differ in the last coding exon (data not shown).

Purified DNA from N2 or lin-26(n156) animals, or DNA from lin-26(mc1) or lin-26(mc2) single dead eggs treated as described by Williams et al. (1992) was amplified by the polymerase chain reaction using Taq Polymerase (Cetus). Following amplification, DNA was run on a low melting point agarose gel, the desired band was excised, reamplified with the same oligonucleotide pair, again gel purified, extracted from agarose with a GeneClean kit (Bio 101) and directly used for DNA sequence determination. The following oligonucleotide pairs were used:

5′TACGGATCCGACGCCGTCTCGGAAAGCTGAAGwith 5′TCTAAGCTTTAATAACCTGATTAAATTTATTATTTG,5′CGTGGATCCATGATTCTTCATGGCTTT with 5′TCTAAGCTTGTGCTGCTCCTCCAGCA, 5′TACGGATCCTCTTCATCATCAACTATCTAC with 5′TACGGATCCACAGAGTGCACACTTGTA, 5′TCTAAGCTTGAAGAGAGAGTCGATCA with 5′TACGGATCCCATCTATGAATAGCTGCTTCGAAC, 5′ATGCTTCCCAGATTATGTGC with 5′CTCGAATGCTGATGAAAGAATGC.

Antisera against GABA were used as described by McIntire et al. (1993). Primary antibodies were revealed with FITC-coupled anti-rabbit antibodies (Cappel) incubated overnight at room temperature. DAPI at 2 μg/ml was included during this second incubation to visualize nuclei (Wood et al., 1988). Animals were mounted on a slide with phenylenediamine in 50% glycerol/50% PBS and observed with a Zeiss Axioplan microscope equipped with fluorescence capabilities.

Ferguson et al. (1987) reported that, in homozygous lin-26(n156) animals, the normally hypodermal Pn.p cells have a neuronal morphology or divide to produce neuronal-like cells, as do their sisters, the neuroblasts Pn.a. We now report that, in lin-26(n156) animals, the Pn.p cells or cells generated by the Pn.p cells can express a neuronal marker normally expressed by descendants of the Pn.a cells, the neurotransmitter GABA (γ-amino-butyric acid). Normally, there are 19 GABAergic neurons in the ventral cord of a wild-type adult animal, 12 of which are descendants of the 12 Pn.a cells (McIntire et al., 1993). By contrast, we found an average of 26±5 GABAergic neurons (n=82) present in the ventral nerve cords of lin-26(n156) animals (Fig. 1). We counted all neurons present between the ventral nerve cord neurons VD2 and VD12 (this region includes neurons derived from the postembryonic ectoblasts P2-P11 and the interspersed DA, DB, DD neurons; Sulston and Horvitz, 1977). Wild-type animals had 57.5±2.5 neurons of which 13±1 were GABAergic (n=54), whereas lin-26(n156) animals had 65.5±11.5 neurons of which 19.5±4.5 were GABAergic (n=82). These observations suggest that most Pn.p cells differentiate as GABAergic neurons and do so without dividing.

Pn.p cells are probably not the only cells that can be transformed to adopt a neural fate in lin-26(n156) animals.

Normally, a single GABA-positive neuron is found posterior to the rectum, the neuron DVB (McIntire et al., 1993). We observed two GABA-positive neurons at that position in two lin-26(n156) animals (n=82), in contrast to the single GABA-positive neuron as expected in all wild-type animals examined (n=54). The additional GABAergic neuron in lin-26(n156) animals could be the cell K.a, as K.a is the sister of DVB and is normally a rectal cell.

The mutation lin-26(n156) is lethal in trans to a deficiency that deletes lin-26 (Ferguson and Horvitz, 1985). We examined such hemizygous lin-26(n156)/Df larvae and found that they had many hypodermal defects (Table 1). C. elegans hypodermal cells can be classified into three broad classes: cells that compose the external epithelium (White, 1988), the excretory system (Nelson and Riddle, 1984) and the rectum (White, 1988). In hemizygous lin-26(n156)/Df animals, all three classes of hypodermal cells appeared to be defective. First, about half of the lin-26(n156)/Df hemizygous animals were dumpy (Table 1; Fig. 2A, B). A dumpy body shape can be a consequence of hypodermal defects, as microfilaments present in the external epithelium are essential for embryo elongation (Priess and Hirsh, 1986) and as some dumpy mutations affect genes encoding cuticle collagens synthesized by the hypodermis (see von Mende et al., 1988; Kramer et al., 1990). Second, degeneration of the hypodermis was observed, particularly in the tail (Table 1; Fig. 2A, B). Third, the hypodermal excretory cell, canal and duct were vacuolated in 60% of hemizygous lin-26(n156)/Df animals (Table 1; Fig. 2C, D). Fourth, the hypo-dermal rectum appeared by Nomarski microscopy to be closed in 24% of hemizygous lin-26(n156)/Df animals (Table 1; Fig. 2E, F). In general, hypodermal cell fates were abnormal but did not appear to be transformed to an alternative fate, as they are in the case of the Pn.p cells. These hypodermal defects seem likely to be the cause of inviability of lin-26(n156)/Df animals. Most of the defects of lin-26(n156)/Df animals were also observed in homozygous lin-26(n156) animals, but at a much lower penetrance (Table 1). This observation suggests that the mutation lin-26(n156) is a weak loss-of-function mutation in a gene that affects all hypodermal cells.

We isolated three new lin-26 alleles, lin-26(mc1), lin-26(mc2) and lin-26(mc4), by a clonal non-complementation screen (see Materials and Methods). All three alleles resulted in embryonic lethality and failed to complement each other and lin-26(n156) for viability. All three new lin-26 alleles appeared to affect hypodermal cells more strongly and probably at an earlier stage of hypodermogenesis than did lin-26(n156). The defects displayed by lin-26(mc1) and lin-26(mc4) embryos are extremely similar to each other and more severe than those displayed by lin-26(mc2) embryos.

In homozygous lin-26(mc1) or lin-26(mc4) embryos, hypo-dermal cells became vacuolated, were difficult to see and reduced in number (Fig. 2G,I), suggesting that these cells had not been generated or that they were dying. For instance, in a wild-type comma-stage embryo there are 21 hypodermal nuclei that correspond to the nuclei of the body hypodermis (Sulston et al., 1983). We observed 21.1±1 such nuclei among 40 control embryos. By contrast, we observed 19.1±7.1 body hypodermal nuclei in lin-26(mc1) comma-stage embryos (n=24) and 17.7±5.7 in lin-26(mc1) terminal-stage embryos (n=30), 19.5±4.5 in lin-26(mc4) comma-stage embryos (n=36) and 16±7 in lin-26(mc4) terminal-stage embryos (n=36). We presume that this reduction in hypoder-mal cells involved a process distinct from programmed cell death (Ellis et al., 1991), as the dominant mutation ced-9(n1950) did not prevent them. (n1950 blocks programmed cell deaths; Hengartner et al., 1992.) In addition, in lin-26(mc1) or lin-26(mc4) embryos, a few rare cells that occupied normal positions of hypodermal cells looked like neurons, suggesting that hypodermal cells might occasionally have adopted a neural fate (data not shown). Hypodermal cells that normally migrate circumferentially around the embryo starting from a dorsal position (Sulston et al., 1983) were not always found at their normal positions but rather were grouped in a dorsolateral position (data not shown). In 26% of lin-26(mc1) embryos (n=249) and 21% of lin-26(mc4) embryos (n=227), internal cells and organs burst from the ventral surface (Fig. 2K,L), presumably because hypodermal cells failed to join together in a ventral position as they should at the end of gastrulation (see Sulston et al., 1983). Finally, most lin-26(mc1) embryos and most lin-26(mc4) embryos did not elongate beyond the comma stage of embryogenesis (Fig. 3), consistent with the observation that microfilaments within the hypodermis are essential for embryo elongation (Priess and Hirsh, 1986). By contrast, non-hypodermal organs appeared normal: lin-26(mc1) and lin-26(mc4) embryos formed a buccal cavity and a shortened but otherwise normal pharynx (Fig. 2H,J,L), an intestine, a gonad primordium and muscles that could twitch (data not shown). Homozygous lin-26(mc2) embryos elongated normally but generally did not hatch (Fig. 3). At the end of embryogenesis, entire cells and droplets were seen floating within the egg shell, suggesting leakage of cells and cytoplasm through holes in the hypodermis (Fig. 2M,N). In addition, the hypodermis and the cuticle, which is synthesized by the hypodermis, appeared rough (Fig. 2M,N). The number and positions of hypodermal nuclei seemed normal.

We suspect that none of these three new lin-26 mutations is a null mutation. Hemizygous lin-26(mc1)/Df embryos failed to elongate and arrested at the lima-bean stage of embryogenesis more frequently than did homozygous lin-26(mc1) embryos (44% versus 23%; see Fig. 3; P≤0.01). Similarly, hemizygous lin-26(mc4)/Df embryos failed to elongate and arrested at the lima-bean stage of embryogenesis more frequently than did homozygous lin-26(mc4) embryos (39% versus 20%, see Fig. 3; P≤0.01). However, these lin-26(mc1)/Df and lin-26(mc4)/Df hemizygotes did not display a phenotype more severe than the most severe phenotype displayed by lin-26(mc1) or lin-26(mc4) homozygotes, arrest at the lima-bean stage. lin-26(mc1) and lin-26(mc4) behaved like a deficiency when het-erozygous with lin-26(mc2), a partial loss-of-function mutation (see Fig. 3). lin-26(mc4) behaved also like a deficiency when heterozygous with lin-26(n156); however, lin-26(mc1) did not behave like a deficiency in this test, as fewer lin-26(n156)/lin-26(mc1) heterozygous embryos hatched than lin-26(n156)/Df hemizygous embryos hatched (Fig. 3). Together, these data suggest that the mutations lin-26(mc1) and lin-26(mc4) eliminate most but not all lin-26 gene function and raise the possibility that the null phenotype of lin-26 could be arrest at the lima bean stage of embryogenesis.

We cloned the gene lin-26 by transformation rescue and localized this gene to an 8 kb restriction fragment (Fig. 4). We identified four abundant RNAs (2.2 kb, 1.5 kb, 1.45 kb and 1.4 kb) and two rare RNAs (3.2 kb and 4.4 kb; these RNAs were not further characterized) transcribed from this region. We identified the 2.2 kb transcript as the lin-26 RNA as described in Materials and Methods. In brief, we introduced frameshift mutations within the coding sequences of each of the four abundant RNAs. Only the mutation in the 2.2 kb RNA abolished the ability of the 8 kb restriction fragment to complement the vulvaless phenotype of lin-26(n156) mutants (Fig. 4C).

The deduced amino acid sequence of the protein encoded by lin-26 contains domains that strongly suggest that this protein is a transcription factor. Specifically, the LIN-26 protein has a serine/threonine-rich amino-terminal region, a glutamine-rich region, two copies of a cysteine/histidine motif (Cys/His motif), and a weakly acid-rich carboxy-terminal region (Fig. 5A,B); all of these types of domains have been associated with transcription factors (Ptashne, 1988). In addition, there is a proline/glutamic acid/serine/threonine-rich region (PEST motif), which is found in proteins with a short half-life (Rogers et al., 1986). The LIN-26 Cys/His motifs might form zinc-fingers, as they are similar to, although distinct from, the TFIIIA zinc-fingers that allow the Xenopus TFIIIA transcription factor to bind nucleic acids (Rhodes and Klug, 1986). In particular, the Cys/His motifs of the LIN-26 protein contain a normally invariant leucine and some residues of the TGEKPY stretch that links two consecutive TFIIIA zinc-fingers (Fig. 5C). Two features of the LIN-26 Cys/His motifs differ from those of canonical TFIIIA-type zinc fingers: the LIN-26 motifs lack a conserved phenylalanine and are somewhat longer than typical, with 17 residues between the second cysteine and the first histidine as opposed to the usual 12 residues. However, some zinc fingers, including two zinc-fingers of TFIIIA itself (Rhodes and Klug, 1986) lack this phenylalanine, and other zinc fingers – such as the second zinc-finger of the mammalian GLI1 protein (Pavletich and Pabo, 1993) – are nearly as long as the LIN-26 Cys/His motifs. Thus, we believe that these motifs are likely to encode TFIIIA-like zinc fingers.

Three lin-26 mutations affect the first putative zinc finger of the LIN-26 protein. The strong embryonic lethal allele lin-26(mc1) changes a glycine to glutamic acid, the weaker embryonic lethal allele lin-26(mc2) changes the adjacent serine to phenylalanine, and the viable allele lin-26(n156) changes the conserved leucine residue to phenylalanine (Fig. 5C). The fourth allele, lin-26(mc4), another strong embryonic lethal allele, is an amber mutation that affects a tryptophan residue located 11 amino acids carboxy terminal to the end of the second zinc finger (Fig. 5A).

The lin-26 gene appears to affect all C. elegans hypodermal cells. The strongest alleles, lin-26(mc1) and lin-26(mc4), caused many hypodermal cells to die during embryogenesis. As a consequence, lin-26(mc1) and lin-26(mc4) homozygous embryos did not elongate and subsequently died. Another embryonic lethal mutation, lin-26(mc2), caused hypodermal cells to fail to enclose the embryo properly. The weakest mutation, lin-26(n156), caused the hypodermal Pn.p cells to adopt a neural fate; in hemizygous lin-26(n156)/Df animals, other hypodermal cells degenerated, although their fates were generally unchanged.

The simplest interpretation of our results is that a complete loss of lin-26 function causes arrest at the lima-bean stage. First, the four lin-26 alleles we have studied result in a recessive phenotype and arose at a frequency typical for loss-of-function mutations (Brenner, 1977; Meneely and Herman, 1979; Greenwald and Horvitz, 1980). Second, two of the mutations we have characterized, lin-26(mc1) and lin-26(mc4), behave as strong loss-of-function alleles. Although a higher percentage of lin-26(mc1)/Df or lin-26(mc4)/Df hemizygous embryos arrested at the lima bean-stage of embryogenesis than did lin-26(mc1) or lin-26(mc4) homozygous embryos, sug-gesting that these mutations are not null alleles, the phenotype of these hemizygous embryos was not more severe than that of the most severely affected homozygous embryos.

Based upon our observations, lin-26 might function in any of three distinct ways: to specify the hypodermal cell fate, to allow the expression of the hypodermal cell fate or to allow the maintenance of the hypodermal cell fate once expressed. That in lin-26 animals some presumptive hypodermal cells express a neuronal fate suggests that lin-26 might in those cells specify the hypodermal fate as opposed to the neuronal fate. That many hypodermal cells die in lin-26(mc1) or lin-26(mc4) mutant embryos also supports the notion that lin-26 might be required for cells to acquire the hypodermal fate. Since differentiated hypodermal cells can degenerate in lin-26(n156)/Df hemizygous animals, lin-26 might also be required to maintain the differentiated state of hypodermal cells.

lin-26 encodes a protein that we presume to be a zinc-finger transcription factor. Two not mutually exclusive models could explain how lin-26 controls hypodermal differentiation. At a biochemical level, lin-26 could encode either a transcriptional activator or a transcriptional repressor. If lin-26 encodes a repressor that acts within the ectoderm, likely targets are genes that specify the neural fate. One argument in favor of a model in which lin-26 acts by preventing neural fates derives from the observation that the 12 Pn.p hypoder-moblasts adopt a neural fate in lin-26(n156) mutants. An alternative hypothesis is that lin-26 acts not by preventing neural fates but rather by promoting non-neural fates. If lin-26 encodes an activator, likely targets are genes that specify the fate of each category of hypodermal cell. In this case, why do the Pn.p cells become neurons in lin-26(n156) mutants? First, lin-26 might act as a repressor of neural-specific genes in the Pn.p cells but not in other hypodermal cells. Alternatively, the lin-26(n156) mutation might confer a novel activity to the LIN-26 protein, causing it to activate the transcription of genes that are not activated by the wild-type LIN-26 protein.

We thank very much Steve McIntire for sharing his observation that lin-26(n156) animals have extra GABAergic neurons. We thank Daniella Rhodes and Grant Jacobs for useful comments about the structure of LIN-26 zinc fingers. We thank Stuart Kim for a sample of his cDNA library, and members of the laboratory of Barbara Meyer for good protocols. Some strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRR) from the NIH. We thank Erik Jorgensen and other members of the Horvitz laboratory for numerous discus-sions. We are very grateful to Cori Bargmann, Michael Basson, Andrew Chisholm, Elisabeth Georges, Angela Giangrande, Erik Jorgensen, Josh Kaplan and Pat Simpson for critical reading of the manuscript. M. L. thanks Bernard Boulay for help with art work. Research at MIT was supported by US Public Health Service grants GM24663 and GM24943 to H. R. H. Research in Strasbourg was supported by a CNRS ATIPE grant to M.L. and funds from the CNRS, INSERM, Centre Hospitalier Universitaire Régional, the Association pour la Recherche sur le Cancer and the Fondation pour la Recherche Médicale. M. L. was supported by Fellowships from the NIH/John Fogarty Center and the Human Frontier Science Program Organiz-ation. H. R. H. is an Investigator of the Howard Hughes Medical Institute.

The accession number within the EMBL databank for the sequence presented in Fig. 5 is Z32673.